ABSTRACT
As modern networks become faster, and the demands for increased bandwidth and low latency become more prevalent, computational resources are unable to keep pace. The problem is the way in which networking is traditionally handled. It is inefficient, not only in terms of CPU utilization, but also in access to memory. Today’s CPU and memory subsystem can handle networks on the order of 1 Gbps, but as network speeds reach 10 Gbps and beyond, problems begin to arise. The crux of the matter is that as networks become faster, so must the CPU in order to process the associated protocols. The rate of increase in CPU clock rate has fallen off dramatically in recent years, while the networking rate has continued to increase. Squeezing ever more cores onto a single CPU and adding more processors to motherboards helps to mitigate the problems, but only to a limited extent.
